Software Engineer- AI/ML, AWS Neuron Distributed Training

AWS Neuron is seeking a talented Software Engineer to join their Machine Learning Applications (ML Apps) team. This role is part of the innovative Annapurna Labs organization, which was acquired by AWS in 2015 and serves as the infrastructure backbone of AWS.

The position focuses on developing and optimizing AWS Neuron, the complete software stack for AWS Inferentia and Trainium cloud-scale machine learning accelerators. You'll be working with cutting-edge ML technologies, including large language models like GPT2 and GPT3, stable diffusion, and Vision Transformers.

As a Software Engineer II, you'll collaborate with chip architects, compiler engineers, and runtime engineers to create sophisticated distributed training solutions. Your responsibilities will include implementing distributed training support in frameworks like PyTorch and TensorFlow, optimizing model performance on AWS Trainium and Inferentia silicon, and working with various ML model families.

The role offers an exciting opportunity to work at the intersection of hardware and software, directly impacting how businesses leverage machine learning at scale. You'll be part of a team that has delivered groundbreaking products like AWS Nitro, ENA, EFA, Graviton, and F1 EC2 Instances.

AWS provides a supportive and inclusive work environment with a strong focus on work-life balance. The company offers comprehensive benefits, mentorship opportunities, and a culture that celebrates diversity through various employee-led affinity groups. You'll have the chance to grow professionally while working on challenging problems that affect millions of users worldwide.

The compensation is competitive, ranging from $129,300 to $223,600 per year, depending on location and experience, plus additional benefits and potential equity. This is an excellent opportunity for someone with strong software development skills and ML knowledge who wants to make a significant impact in the cloud computing and machine learning space.